pirl-unc / hitlist / 24421411603
80%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 18
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 43.724% (+9.3%) from 34.44%
24421411603

push

github

web-flow 
v1.7.0: Split MS and binding assays into separate parquet indexes (#48)

* v1.7.0: Split MS and binding assays into separate parquet indexes

hitlist data build now writes TWO parallel parquet files:
- observations.parquet — MS-eluted immunopeptidome (IEDB+CEDAR+supplementary)
- binding.parquet — binding-assay rows (microarray, refolding, MEDi,
  Positive-High/Intermediate/Low quantitative tiers)

The two indexes share the same schema plus gene annotations but are
never mixed. Supplementary data is MS-only. Each index is queryable
independently:

  from hitlist.observations import load_observations, load_binding
  hitlist export observations [filters] -o ms.csv
  hitlist export binding      [filters] -o binding.csv

Binding export drops MS-specific filters (--mono-allelic,
--instrument-type, --acquisition-mode) and does not apply the
sample-metadata join (binding rows have no MS sample context).

Bundles the v1.6.1 --serotype filter work: accepts locus-specific
names (A24, B57, DR15) or public epitopes (Bw4, Bw6, C1, C2);
canonical serotype prefers locus-specific over public epitope
(A*24:02 → HLA-A24, not HLA-Bw4); new serotypes column is a
semicolon-joined multi-valued list; post-filter over the full list.

Mappings sidecar now covers the union of peptides from both indexes.
Cache invalidation requires both parquets to be present; older
installs rebuild once on upgrade.

Also documents the PMID-level mono-allelic curation model: is_monoallelic
can be True with empty mhc_restriction for supplementary contaminant
rows under a mono_allelic_host override. Readers who need a strict
"mono-allelic AND has allele" subset must post-filter.

* Fix is_monoallelic: require per-row allele, skip validation samples

Treat mono-allelic as a sample-level claim rather than a PMID-level
one.  The previous PMID-level `mono_allelic_host` override blanket-
flagged every row in an overridden PMID, which was wrong in two
real curated cases:

1. PMID 30315122 (Faridi 2018) ... (continued)

1160 of 2653 relevant lines covered (43.72%)

0.44 hits per line

Coverage Regressions

Lines Coverage ∆ File
207
40.27
 40.27% builder.py
182
34.84
 11.7% downloads.py
163
0.0
 0.0% cli.py
76
23.18
 0.86% scanner.py
44
83.39
 9.0% curation.py
35
85.55
 9.82% export.py
18
71.74
 -2.7% proteome.py
8
90.0
 32.86% observations.py
Jobs
ID Job ID Ran Files Coverage
1 24421411603.1 18
43.72
 GitHub Action Run
Source Files on build 24421411603